You might be doing basic stuff, like copying and pasting text on spread sheets, you might be running large scale automated platforms with sophisticated algorithms, or somewhere in between. In any case your job is all about working with data.
As a data scientist you don't spend most of your time producing cool visualizations or finding great insights. The majority of your time is spent wrangling with URLs, figuring out how to stitch together two tables, hoping that the dates, won't break, without you knowing, or trying to generate the next 124,538 keywords for an upcoming campaign, by the end of the week!
advertools is a Python package that can hopefully make that part of your job a little easier.
python3 -m pip install advertoolsIt's very easy to learn how to use advertools. There are two main reasons for that.
First, it is essentially a set of independent functions that you can easily learn and use. There are no special data structures, or additional learning that you need. With basic Python, and an understanding of the tasks that these functions help with, you should be able to pick it up fairly easily. In other words, if you know how to use an Excel formula, you can easily use any advertools function.
The second reason is that advertools follows the UNIX philosophy in its design and approach. Here is one of the various summaries of the UNIX philosophy by Doug McIlroy:
Write programs that do one thing and do it well. Write programs to work together. Write programs to handle text streams, because that is a universal interface.
Let's see how advertools follows that:
Do one thing and do it well: Each function in advertools aims for that. There is a function that just extracts hashtags from a text list, another one to crawl websites, one to test which URLs are blocked by robots.txt files, and one for downloading XML sitemaps. Although they are designed to work together as a full pipeline, they can be run independently in whichever combination or sequence you want.
Write programs to work together: Independence does not mean they are unrelated. The workflows are designed to aid the online marketing practitioner in various steps for understanding websites, SEO analysis, creating SEM campaigns and others.
Programs to handle text streams because that is a universal interface: In Data Science the most used data structure that can be considered “universal” is the DataFrame. So, most functions return either a DataFrame or a file that can be read into one. Once you have it, you have the full power of all other tools like pandas for further manipulating the data, Plotly for visualization, or any machine learning library that can more easily handle tabular data.
This way it is kept modular as well as flexible and integrated. As a next step most of these functions are being converted to no-code interactive apps for non-coders, and taking them to the next level.
The most important thing to achieve in SEM is a proper mapping between the three main elements of a search campaign
Keywords (the intention) -> Ads (your promise) -> Landing Pages (your delivery of the promise) Once you have this done, you can focus on management and analysis. More importantly, once you know that you can set this up in an easy way, you know you can focus on more strategic issues. In practical terms you need two main tables to get started:
- Keywords: You can generate keywords (note I didn't say research) with the kw_generate function.
- Ads: There are two approaches that you can use:
- Bottom-up: You can create text ads for a large number of products by simple replacement of product names, and providing a placeholder in case your text is too long. Check out the ad_create function for more details.
- Top-down: Sometimes you have a long description text that you want to split into headlines, descriptions and whatever slots you want to split them into. ad_from_string helps you accomplish that.
- Tutorials and additional resources
- Get started with Data Science for Digital Marketing and SEO/SEM
- Setting a full SEM campaign for DataCamp's website tutorial
- Project to practice generating SEM keywords with Python on DataCamp
- Setting up SEM campaigns on a large scale tutorial on SEMrush
- Visual tool to generate keywords online based on the kw_generate function
Probably the most comprehensive online marketing area that is both technical (crawling, indexing, rendering, redirects, etc.) and non-technical (content creation, link building, outreach, etc.). Here are some tools that can help with your SEO
- SEO crawler:
A generic SEO crawler that can be customized, built with Scrapy, & with several
features:
- Standard SEO elements extracted by default (title, header tags, body text, status code, response and request headers, etc.)
- CSS and XPath selectors: You probably have more specific needs in mind, so you can easily pass any selectors to be extracted in addition to the standard elements being extracted
- Custom settings: full access to Scrapy's settings, allowing you to better control the crawling behavior (set custom headers, user agent, stop spider after x pages, seconds, megabytes, save crawl logs, run jobs at intervals where you can stop and resume your crawls, which is ideal for large crawls or for continuous monitoring, and many more options)
- Following links: option to only crawl a set of specified pages or to follow and discover all pages through links
- robots.txt downloader A simple downloader of robots.txt files in a DataFrame format, so you can keep track of changes across crawls if any, and check the rules, sitemaps, etc.
- XML Sitemaps downloader / parser An essential part of any SEO analysis is to check XML sitemaps. This is a simple function with which you can download one or more sitemaps (by providing the URL for a robots.txt file, a sitemap file, or a sitemap index
- SERP importer and parser for Google & YouTube Connect to Google's API and get the search data you want. Multiple search parameters supported, all in one function call, and all results returned in a DataFrame
- Tutorials and additional resources
- A visual tool built with the
serp_googfunction to get SERP rankings on Google - A tutorial on analyzing SERPs on a large scale with Python on SEMrush
- SERP datasets on Kaggle for practicing on different industries and use cases
- SERP notebooks on Kaggle some examples on how you might tackle such data
- Content Analysis with XML Sitemaps and Python
- XML dataset examples: news sites, Turkish news sites, Bloomberg news
- A visual tool built with the
URLs, page titles, tweets, video descriptions, comments, hashtags are some
examples of the types of text we deal with. advertools provides a few
options for text analysis
- Word frequency
Counting words in a text list is one of the most basic and important tasks in
text mining. What is also important is counting those words by taking in
consideration their relative weights in the dataset.
word_frequencydoes just that. - URL Analysis
We all have to handle many thousands of URLs in reports, crawls, social media
extracts, XML sitemaps and so on.
url_to_dfconverts your URLs into easily readable DataFrames. - Emoji
Produced with one click, extremely expressive, highly diverse (3k+ emoji),
and very popular, it's important to capture what people are trying to communicate
with emoji. Extracting emoji, get their names, groups, and sub-groups is
possible. The full emoji database is also available for convenience, as well
as an
emoji_searchfunction in case you want some ideas for your next social media or any kind of communication - extract_ functions The text that we deal with contains many elements and entities that have their own special meaning and usage. There is a group of convenience functions to help in extracting and getting basic statistics about structured entities in text; emoji, hashtags, mentions, currency, numbers, URLs, questions and more. You can also provide a special regex for your own needs.
- Stopwords A list of stopwords in forty different languages to help in text analysis.
- Tutorial on DataCamp for creating the
word_frequencyfunction and explaining the importance of the difference between absolute and weighted word frequency - Text Analysis for Online Marketers An introductory article on SEMrush
In addition to the text analysis techniques provided, you can also connect to
the Twitter and YouTube data APIs. The main benefits of using advertools
for this:
- Handles pagination and request limits: typically every API has a limited number of results that it returns. You have to handle pagination when you need more than the limit per request, which you typically do. This is handled by default
- DataFrame results: APIs send you back data in a formats that need to be parsed and cleaned so you can more easily start your analysis. This is also handled automatically
- Multiple requests: in YouTube's case you might want to request data for the same query across several countries, languages, channels, etc. You can specify them all in one request and get the product of all the requests in one response
- Tutorials and additional resources
- A visual tool to check what is trending on Twitter for all available locations
- A Twitter data analysis dashboard with many options
- How to use the Twitter data API with Python
- Extracting entities from social media posts tutorial on Kaggle
- Analyzing 131k tweets by European Football clubs tutorial on Kaggle
- An overview of the YouTube data API with Python
Function names mostly start with the object you are working on, so you can use autocomplete to discover other options:
kw_: for keywords-related functionsad_: for ad-related functionsurl_: URL tracking and generationextract_: for extracting entities from social media posts (mentions, hashtags, emoji, etc.)emoji_: emoji related functions and objectstwitter: a module for querying the Twitter API and getting results in a DataFrameyoutube: a module for querying the YouTube Data API and getting results in a DataFramecrawlytics: a module for analyzing crawl data (compare, links, redirects, and more)serp_: get search engine results pages in a DataFrame, currently available: Google and YouTubecrawl: a function you will probably use a lot if you do SEO*_to_df: a set of convenience functions for converting to DataFrames
(log files, XML sitemaps, robots.txt files, and lists of URLs)